The First Time Google Alerted Me to Leave for an Appointment

The first time Google alerted me to leave for an appointment, it was one of those “Aha!” moments. I thought to myself, “That’s amazing!” as I realized it was smart enough to check my calendar, learn where and at what time my appointment was, calculate the time I needed to leave to arrive on time, and then sent me an alert to depart.

But how quickly we adapt and take much of this for granted. Now there are numerous apps that make decisions and are trying to think for us. They monitor our email, calendar, and location and, using artificial or some other sort of intelligence, take action we’d normally do on our own.

TripIt will detect when you’re emailed an itinerary from an airline and add the flights to your calendar. Waze will determine when a traffic jam occurs and route you around it. Google will determine when an appointment invite is emailed to you and put it on your calendar. So many apps now have this level of intelligence to create an action without our intervention.

I’m finding, however, as this occurs more frequently, there are numerous errors that occur. I need to intervene more often and make corrections because the assumptions these apps make are usually based on fairly primitive logic. Not every mention in an email is meant to be actionable. Not every itinerary is yours and needs to be scheduled. While the apps try to be helpful, there are many times they get it wrong; perhaps 25% of the time.

As an example, a misaddressed email sent to me contained a receipt for train tickets bought in the UK and TripIt added it to my calendar. I received a Google Now alert to leave for a meeting, even though it was a phone meeting, whose invitation was emailed to me and contained an address. Sometimes I find an action is created twice because more than one app is monitoring the same thing. A plane flight will often show up twice on my calendar, put there by both TripIt and Google. And now the Calendar on iOS 10 is starting to look for events to add automatically. Yes, I can go back to each of the apps’ settings and turn some off or others on. But as more apps do this without asking, it’s time-consuming. Clearly, the simple logic that awed me the first time may not be sufficient without becoming much more intelligent.

It’s not that big a deal right now since we can make these manual corrections and learn which apps work well and which do not. But it shows there are some serious challenges the software has to overcome to work reliably enough so we won’t turn it off and avoid using it all together. Some apps will need to be more careful or less aggressive than others. Some will be very aggressive and make many more assumptions. The burden is on us to filter out what doesn’t work and use what does.

While these new capabilities are designed to simplify our lives, for now, they seem to be requiring more time to manage because of the unintended consequences.

What needs to be done? The software needs to learn our habits and preferences and tailor their actions on an individual basis. They need to query us sometimes before taking action. When Waze sends us on a circuitous local route with numerous turns through a sketchy neighborhood to avoid freeway congestion, it needs to inform us what it’s doing, what to expect, and ask us if we really want to do it.

This is an exciting area with lots of potential to think for us, but we’re just in its infancy and there’s a lot more room for improvement.

Published by

Phil Baker

Phil Baker is a product development expert, author, and journalist covering consumer technology. He is the co-author with Neil Young of the forthcoming book, “To Feel the Music,” and the author of “From Concept to Consumer.” He’s a former columnist for the San Diego Transcript, and founder of Techsperts, Inc. You can follow him at www.bakerontech.com.

15 thoughts on “The First Time Google Alerted Me to Leave for an Appointment”

  1. AS far as I know, the automatic “add event to calendar from email” feature in iOS 10 requires action initiated from the user. It’s hence easier to make sure that no mistakes creep in, and to prevent double entries.

    The same is true for the intelligent folder management feature, which is triggered by a “move to folder” action and uses AI to predict which folder you want to put the message into.

    AI tasks are not created equal. Some have a high tolerance for failure (like a photo categorisation) and others must be 100% correct (like scheduling). For the latter, AI is basically an enhancement to the user interface, and the companies that understand UI/UX have the upper hand.

  2. “…there are many times they get it wrong; perhaps 25% of the time”

    That’s about how much it gets it correct for me.

    Joe

  3. For the “intelligent” software to get it wrong 25% of the time would be intolerable for me. i would rather turn the feature off than have to correct so many mistakes. I doubt I am alone in this.

    Take voice dictation: on IOS, it works really well, and is much faster than typing, but it gets occasional words wrong and makes horrible guesses when faced with words it doesn’t know. Overall error rate is probably only 15%, but that is enough to ensure that I never use dictation instead of typing unless I am in an incredible hurry or have something incredibly simple to dictate or am trying to input a ton of text and want a break from typing it all.

    1. Regarding voice dictation, due to sheer volume 1-2% would be too high for my tolerance. And then need to correct on a touchscreen….

  4. Google has been doing what the author discovered for years. Siri has always been too stupid for me. google now has been great for me. I just add something to my calendar with the start and end times and the location and reminders to leave are posted for me with correct estimates for travel time with traffic taken into account. I do not use more than one app other than google calendar. I use google mail as my mail app. The biggest problem is folks try to use an app for every single individual thing…they fall into the app vendors swansong. I simply use the google apps for just about everything actionable. Now it helps I have been on the google platform since android version 2.0. Get on a platform and give it time to learn you. yes it gets uncanny how well it can anticipate you after a while. However…for me…Google has the digital assistant wars won by far over apples IOS. Others..YMMV. Keep in mind though as you use multiple apps they all have to learn you and few companies have the abilities Google does..:)

  5. This discussion is highlighting what bothers me a lot about Artificial Intelligence, which I am inclined to rename Artificial Stupidity, as in the stupidity of artificial minds.

    1. Computers ‘learn’ from data. It supposedly ‘learns’ your preferences by recognizing your patterns of action. The more you do certain things, certain ways, in a certain sequence, the faster it recognizes a pattern. In short, what it learns and thus ‘predicts’ first are the most common events. You throw it a new action or sequence of actions, then it’s lost. At the margin of its ‘intelligence’ when it is encountering a new pattern, computers will be making mistakes.

    Now the thing about the world that I’ve noticed is that the bigger the mishap is, the rarer it occurs. Really big catastrophes come few and far between. So if we start relying on AI to make life easier for us, its scope of competence will start out with the more common, less problematic events and this scope expands as it gathers more data and recognizes more patterns. Imagine years pass by and AI learns so much about all sorts of patterns that they don’t make mistakes anymore ON THE COMMON EASY THINGS. What that means is that we’re going to not only rely, but trust, computers to run things for us until that rare catastrophic event arises and who will be paying attention then? Who will have the expertise to address the disaster then? It’s the same hand-off problem with self-driving cars at a higher order of magnitude: The computer hands off control to of the car to an inattentive human at the moment of maximum confusion.

    2. There is a famous, much reproduced experiment in psychology (that field is notorious for irreprucible experimental results) where they show that humans, down to 2 years old even, understand the concept of ‘people other than myself have their own thoughts and own sets of knowledge’ and, more importantly, can make accurate guesses of what another person is thinking of and knows.

    Earlier this week, news came out that this experiment was run on bonobos and chimpanzees and the result is that they also have this ability. Here…

    https://today.duke.edu/2016/10/falsebeliefs

    This skill of ‘reading another persons mind’ is a critical component of human intelligence. It is indispensable for humor, satire, empathy, kindness, leadership, poetry, debate, etc. It’s what allows two people to come out of a cinema and start communicating through allusion to specific phrases and scenes they just saw. I’d dare say a computer will never pass a Turing test because it doesn’t have this skill. The lack of this ability is another reason computers, when interacting directly with individuals, will always be making stupid mistakes on the margins of its set of recognized patterns.

    The upshot of this Stupidity of Artificial Minds, is that as AI becomes woven deeper into our daily lives, humans will actually dumb themselves and the world down to a level that makes it less likely that AI driven programs will make mistakes. It’s already going on. We are simplifying our speech patterns so that the homonculus in our smartphones can understand us better. When self-driven cars go mainstream, I expect there will be a push to restrict the range of actions that human-driven cars can do all in the name of safety because the roads is the first large-scale setting where the problem of rare, previously unobserved patterns will have lethal outcomes.

    1. I understand your thoughts on whether AI can become truly human. However, I’m not sure if this is the point.

      The way I think of it is that you can be stupid but still be quite helpful. Computers in general have been quite stupid for the past half century, but still they’ve been immensely helpful. In my opinion, this is what AI should aim for. To be helpful, and not necessarily intelligent.

      The other way to thing of it is, it’s not the intelligence of the algorithms that matter most. More than that, it’s providing a good UI so that humans can easily deal with stupidity, and benefit when it gets things right.

      1. I fully agree with your conception of what AI can do, and should be expected to do. What I am concerned about is that a lot of people think AI can go beyond that and actually try to build and deploy AI systems that try to take over decision-making that should be left for humans, fallible as we are. And the thing is, such a system will work. . . until it doesn’t. And there is no way of predicting where it will fail because if you could then you would have fixed that ahead of time.

        ——-

        “…it’s providing a good UI so that humans can easily deal with stupidity, and benefit when it gets things right.”

        Absolutely. How does a dictation app make a distinction between text for transcription and a command about the text (without reserving words to clue in the app)? A stenographer easily makes this distinction because she has the ability to infer intent. “Check that, say Mrs. Jones rather than Cleopatra.”

  6. The CBD SEO Agency delivers top-notch services quest of businesses in the thriving CBD industry. With a vital nearer to Search Appliance Optimization (SEO), they cbd seo agency outrank in enhancing online visibility and driving constitutional traffic. Their expertise lies in tailoring SEO strategies specifically representing CBD companies, navigating the together challenges of this place market. Middle of wide keyword enquiry, content optimization, and link-building tactics, they effectively encouragement search rankings, ensuring clients experience out amidst competition. Their group’s pledge to staying updated with enterprise trends and search motor algorithms ensures a emphatic and operative approach. The CBD SEO Action’s commitment to transparency and shopper communication fosters keeping and reliability. Total, their specialized services dance attendance on to the noticeable needs of CBD businesses, making them a valuable partner in navigating the digital aspect within this competitive market.

  7. I gave cbd balm for muscles a whack at payment the cardinal adjust, and I’m amazed! They tasted great and provided a sense of calmness and relaxation. My emphasis melted away, and I slept well-advised too. These gummies are a game-changer an eye to me, and I highly commend them to anyone seeking unconstrained worry relief and well-advised sleep.

Leave a Reply

Your email address will not be published. Required fields are marked *